cuffdiff gtf input from cufflinks?

[PFS]

hi,

I would like to estimate difference in expression using cuffdiff. The program requires a GTF file.

If I provide a standard annotation file (such as from UCSC), could I just skip running cufflinks?

On the other hand, if I do want to provide a GTF from cufflinks, which GTF do I provide? There are two samples to compare, and cufflinks must be run separately for each of those, resulting in two GTF trascript files.

Any suggestion?

[Wallysb01]

Yes, you can skip cufflinks when running cuffdiff. It will give you all the same types of readouts (FPKMs, and confidence scores) for all your samples individually, plus confidence calls on the differences between the samples.

As for the second question, you could provide either, but then it could bias your results towards the sample that GTF file came from. What you could do, is run cufflinks as if all your samples are replicates to create a new GTF for the composite, then use that GTF in cuffdiff.

I hear in the future cufflinks will have an option where you can input a GTF and then cufflinks will extend and create new annotations to that GTF. It will be nice when that happens.